Research On Fault Diagnosis Method Of High-speed Train Bogie Based On Local Characteristic Scale Decomposition

As the core component of high-speed train,bogie not only plays the role of traction and bearing,but also is one of the keys to ensure the safe operation of train.By analyzing the vibration signal recorded in the sensor,the fault type can be determined,which is of great significance for the safe and stable operation of the train.However,in the process of train running,the vibration signals show strong nonlinear and non-stationary characteristics.How to extract useful fault characteristics from complex vibration information is a challenging task in high-speed train bogie fault diagnosis.As a signal processing method proposed in recent years,local characteristic scale decomposition method has certain advantages in the precision of signal decomposition,decomposition speed and theoretical completeness.In view of the advantages of local feature decomposition and the nonlinear and non-stationary characteristics of high-speed train vibration signals,this paper adopts the local feature scale decomposition method to study high-speed train fault vibration signals.The main contents are as follows:In this paper,the local characteristic scale decomposition method is studied,and it is compared with the empirical mode decomposition method,which verifies the advantages of the local characteristic scale decomposition method compared with the traditional signal processing methods in decomposition accuracy,decomposition speed and end effect.In order to solve the problem of end effect in local feature scale,an adaptive waveform matching extension method based on cosine similarity is proposed,which solves the problem of matching wavelet step size determination that may appear in the traditional endpoint extension method.The advantages of the proposed method are verified by comparing it with the mirror extension method.In the process of train operation,it is inevitable to receive interference signals,and noise signals will cause the mode aliasing problem of local feature scale decomposition method.In order to ensure the reliability of local feature scale decomposition of train vibration signals,this paper adopts noise-assisted signal processing method to suppress mode aliasing problem.The partial ensemble local feature scale decomposition is applied to the processing of highspeed train fault vibration signals and compared with the complementary set empirical mode decomposition method.The vibration characteristics of the vehicle were studied under the single fault condition with 8 anti-snake shock absorber faults at the operating speed of200km/h,the compound fault condition with 8 kinds of mixed faults of anti-snake shock absorber and lateral shock absorber,and the vibration characteristics of the vehicle with 7kinds of vertical steel spring stiffness degradation.The full vector theory can realize the information fusion of homologous two-channel signals through Fourier transform.In this paper,PELCD method is extended to the level of binary signal processing to solve the problem that homologous two-channel signals may have component scale inconsistency and cannot be fully fused during information fusion.On the other hand,considering the characteristics of signal changes with time,sample entropy and total vector theory are combined to extract train fault features.By comparing with sample entropy feature extraction method without full vector fusion in 3D feature space,the superiority of the fault feature extraction method based on full vector fusion is demonstrated.As for the selection of classifier,considering the relatively small sample data collected in this paper,support vector machine is selected as the classifier for fault diagnosis and grey Wolf optimization algorithm is adopted to optimize the penalty parameters and kernel parameters of support vector machine.The eigenvalue of total vector fusion sample entropy was used as the training sample to identify the single fault condition,compound fault condition and performance degradation fault condition of high-speed train respectively.The experimental results show that the GWO-SVM method with full vector fusion can effectively improve the train fault recognition rate and reduce the time required for training samples.